CN103164962B - Mountain road sharp turn section real-time vehicle speed early warning method - Google Patents

Abstract

The invention particularly discloses a mountain road sharp turn section real-time vehicle speed early warning method. The mountain road sharp turn section real-time vehicle speed early warning method comprises steps as below: step one: arranging basic information collection devices on vehicles and sharp turn roads, collecting information of roads and vehicles in real time; step two: constructing a calculation controller of road attachment coefficients and rolling resistance coefficients based on a BP neural network in a computer; step three: inputting each parameter collected in step one into the BP neural network controller for processing, outputting road attachment coefficients and rolling resistance coefficients of the vehicles at a curve position of the road; step four: calculating distances between the vehicles; step five: calculating the safety distances between the vehicles in the computer; step six: comparing the safety distance of the vehicles and the distances between the vehicles, then carrying out alarming for vehicles exceeding a limit speed.

Description

The real-time vehicle speed prewarning method in section, a kind of mountain highway sharp turn

Technical field

The present invention relates to a kind of vehicle speed prewarning system, specifically can ensure the vehicle speed prewarning method of section, mountain highway sharp turn traffic safety.

Background technology

Vehicle speed prewarning system is a critical facility that ensures highway traffic safety, its Main Function is when there is accident potential, to make driver slow down in time, thereby the generation avoiding traffic accident, particularly when road alignment condition and weather conditions are all poor, in conjunction with meteorological, transportation condition in real time, vehicle are carried out to vehicle speed prewarning and seem particularly important.

The expressway bend vehicle speed prewarning systematic research of the paper of the 12nd phase in 2009 of periodical Shandong Jiaotong University journal-based on image processing techniques in addition; The design of blind area, Northeast Agricultural University's Master of engineering paper-hill path bend vehicle meeting prompt system, the design of the mountain area bend intelligence meeting prompt system of the 6th phases in 2011 of periodical infotech; A automobile cornering meeting Radar Design based on Ling Yang 61 single-chip microcomputers of the 6th phases in 2006 of periodical TV tech, above-mentioned prior art is not considered DIFFERENT METEOROLOGICAL CONDITIONS and the impact of highway layout parameter on vehicle running state mostly, cannot build early warning scheme according to the real time running state of vehicle, with combination Real-time Road, traffic and the meteorological condition of the present invention's proposition, according to the vehicle method difference that transport condition carries out early warning separately.

Whether the different transport conditions when patent curve barrier early warning system (CN200820122886) early warning scheme does not consider that vehicle enters bend have the necessity to its early warning in addition; Patent highway bend vehicle meeting prior-warning device (CN200920063304) mainly for be that meeting is reminded, do not relate to the potential safety hazard between vehicle in the same way; Bend overspeed warning device (CN201120507735) based on video image identification is to carry out a bend vehicle speed prewarning based on video image identification, and this device is poor for applicability under the condition of low visibility.

Summary of the invention

The shortcoming existing for solving prior art, the present invention proposes the real-time vehicle speed prewarning method in section, a kind of mountain highway sharp turn.

The present invention is achieved through the following technical solutions, and concrete steps are as follows:

Step 1. is laid basic information collection device on vehicle and zig zag highway, gathers road and information of vehicles on the spot, and described basic information collection device comprises on-vehicle information harvester and road traffic pick-up unit;

Step 2 builds the computing controller of road-adhesion coefficient and coefficient of rolling resistance in truck-mounted computer based on BP neural network;

The input information BP nerve network controller that step 3 gathers on-vehicle information harvester in step 1 is processed, export vehicle at road-adhesion coefficient and the coefficient of rolling resistance of sharp turn, and transfer to trackside microwave read-write antenna (RSU) equipment and then transfer to center-controlling computer by vehicle-carried microwave communication apparatus (OBU);

The front truck speed of a motor vehicle that step 4 center-controlling computer detects according to road traffic pick-up unit, front truck are laid spacing and the rear car speed of a motor vehicle, rear car current location calculating vehicle actual pitch and are differentiated the current motion state of front truck by time, the road traffic pick-up unit of each road traffic pick-up unit;

The signal of the real-time speed of a motor vehicle of rear car that step 5 center-controlling computer detects according to road traffic pick-up unit, step (3) output and the in advance current section highway layout parameter of storage, obtain the safe distance between vehicle;

Safe distance and the spaces of vehicles of step 6 pair vehicle compare, if spaces of vehicles is less than safe distance, the rear car speed of a motor vehicle are reported to the police, if spaces of vehicles is greater than safe distance, get back to the signal that step 1 gathers next car.

In described step 1, on-vehicle information harvester is by infrared temperature, humidity sensor being set at automobile chassis and saw sensor being set on tire, and collection vehicle enters before highway zig zag within the scope of 20-50m the tire pressure of temperature, humidity and the tire on road surface, deflection, contact area, slippage situation over time; Road traffic pick-up unit mainly gathers the speed of a motor vehicle, the volume of traffic and vehicle and passes through temporal information.Temperature, humidity sensor and saw sensor transfer to carried-on-vehicle computer system by CAN bus and Wireless microwave communication modes by image data respectively, and road traffic pick-up unit is realized and being communicated by letter with center-controlling computer by Optical Fiber Transmission.

Road traffic pick-up unit method to set up on described zig zag highway: from zig zag starting point, considering under highway layout speed of a motor vehicle condition, weather and pavement conditions vehicle safe braking distance empirical value when better of take is normative reference, every this distance, establish one and be subject to weather and the less road traffic pick-up unit of environmental impact, can adopt coil road traffic pick-up unit or microwave road traffic pick-up unit, to realize real-time detection and the transmission of the volume of traffic and vehicle speed data on road;

In described step 2, the process of establishing of computing controller is: using the Various types of data value of on-vehicle information harvester Real-time Collection as input, using road-adhesion coefficient and coefficient of rolling resistance as output, chamber test data and existing correlation parameter historical data are carried out sample training by experiment, training error is reduced to predetermined threshold or reaches frequency of training, to obtain optimal neural network controller;

In described step 3, the nerve network controller that the data input step 2 that when vehicle is entered to zig zag, on-vehicle information harvester gathers builds, try to achieve vehicle at road-adhesion coefficient and the coefficient of rolling resistance of this sharp turn, and utilize Dedicated Short Range Communications, technology (DSRC) transfer to trackside microwave read-write antenna (RSU) equipment and then transfer to center-controlling computer by vehicle-carried microwave communication apparatus (OBU);

In the calculating of described step 4 vehicle headway, when there being vehicle to enter successively after zig zag, each vehicle of road traffic detection device records in zig zag is by time, the speed of a motor vehicle and this link traffic flow, when a upper road traffic pick-up unit detects the correlation parameter of rear car, the time according to road traffic pick-up unit setting space, front truck by this pick-up unit, the speed of a motor vehicle and now whether by next pick-up unit, calculate vehicle headway and judge the transport condition of front truck;

Described step 5 safe distance solve, center-controlling computer is coefficient of road adhesion and the rolling resistance coefficient at this sharp turn according to the definite rear car of step 3, front truck and the rear car speed of a motor vehicle detecting according to step 4 and store in advance Ci highway zig zag Alignment Design parameter, calculates the safe distance in two workshops on this basis;

Speed of a motor vehicle warning implementation in described step 6: before each the road traffic pick-up unit arranging, buzzing warning device is set in zig zag, when not meeting safe distance requirement by above-mentioned calculative determination rear car and leading vehicle distance, when rear car is travelled to warning device place, it is sent to speed of a motor vehicle alarm.

Take a sudden turn in described step 1 setting of road traffic pick-up unit on highway: from zig zag starting point, consider highway layout vehicle speed value, weather and the pavement conditions average safe stopping distance value of vehicle when better of take is that spacing normative reference is arranged road traffic pick-up unit, in order to make road traffic pick-up unit when work as far as possible not affected by environment, select coil road traffic pick-up unit or microwave road traffic pick-up unit.

The computing formula of vehicle safe braking distance is:

$D=\frac{V^2}{254(\mathrm{\φ}+{\mathrm{\ψ}}_1)}$

In formula, D is vehicle safe braking distance, and V is the average speed of typical vehicle when entering this zig zag on road, and ψ ₁wei weather and pavement conditions attachment coefficient and the road resistance coefficient between road surface and tire, wherein ψ when better ₁=f+i _z, f is coefficient of rolling resistance, i _zfor road longitudinal grade.

In described step 2, computing controller is divided into three-decker, i.e. input layer, and hidden layer and output layer, concrete structure is as Fig. 1.

Ground floor is input layer, and input value is expressed as x={x ₁, x ₂... x _m, wherein, M represents the number of input parameter, M is more than or equal to 1 integer, x ₁, x ₂... x _mrepresent respectively vehicle, the speed of a motor vehicle, tire pressure, deflection, contact area, slippage, pavement temperature and humidity; x _jrepresent x ₁, x ₂... x _m(j=1,2, any one parameter in 3...M), the i.e. arbitrary parameter of input layer;

The second layer is hidden layer, the arbitrary node i of hidden layer be input as net _i

${\mathrm{net}}_i=\underset{j=1}{\overset{M}{\mathrm{\Σ}}}{w}_{\mathrm{ij}}{x}_j+{\mathrm{\θ}}_i,$

I node is output as y _i

$y_i=\mathrm{\φ}\left({\mathrm{net}}_i\right)=\mathrm{\φ}(\underset{j=1}{\overset{M}{\mathrm{\Σ}}}{w}_{\mathrm{ij}}{x}_j+{\mathrm{\θ}}_i)$

Wherein

W _ijrepresent that j node of input layer is to the weights of i node of hidden layer, wherein j and i represent respectively the numbering of input layer and the arbitrary node of hidden layer, j=1,2,3...M, i=1,2,3...q, q is more than or equal to 1 natural number, and M and q are respectively the node number of input layer and hidden layer;

θ _ithe threshold value that represents i node of hidden layer;

φ (x) represents the excitation function of hidden layer;

The 3rd layer is output layer, finally obtains coefficient of road adhesion and coefficient of rolling resistance value, k node of output layer be input as net _k

${\mathrm{net}}_k=\underset{i=1}{\overset{q}{\mathrm{\Σ}}}{w}_{\mathrm{ki}}{y}_i+a_k=\underset{i=1}{\overset{q}{\mathrm{\Σ}}}{w}_{\mathrm{ki}}\mathrm{\φ}(\underset{j=1}{\overset{M}{\mathrm{\Σ}}}{w}_{\mathrm{ij}}{x}_j+{\mathrm{\θ}}_i)+a_k$

K node is output as o _k

$o_k=\mathrm{\ψ}\left({\mathrm{net}}_k\right)=\mathrm{\ψ}(\underset{i=1}{\overset{q}{\mathrm{\Σ}}}{w}_{\mathrm{ki}}{y}_i+a_k)=\mathrm{\ψ}(\underset{i=1}{\overset{q}{\mathrm{\Σ}}}{w}_{\mathrm{ki}}\mathrm{\φ}(\underset{j=1}{\overset{M}{\mathrm{\Σ}}}{w}_{\mathrm{ij}}{x}_j+{\mathrm{\θ}}_i)+a_k)$

Wherein

W _kirepresent that the arbitrary node i of hidden layer is to the weights of the arbitrary node k of output layer, k is the arbitrary node serial number of output layer, k=1, and 2,3...L, L is more than or equal to 1 natural number;

A _kthe threshold value that represents k node of output layer;

Ψ (net _k) represent the excitation function of output layer;

O _krepresent the output of k node of output layer;

During sample training, according to the backpropagation of error, first by output layer, start successively to calculate each layer of neuronic output error, then according to error gradient descent method, regulate weights and the threshold value of each layer, make the final output of amended network can approach expectation value;

Input P learning sample, use x ¹, x ²... x ^p, represent; For P sample x ^pquadratic form error rule function be E _p: P is the natural number being more than or equal to;

$E_p=\frac{1}{2}\underset{k=1}{\overset{L}{\mathrm{\Σ}}}{({T_k}^p-{O_k}^p)}^2$

T wherein _kit is the desired output of k node; o _krepresent the output of k node of output layer; T _k ^pand O _k ^pthe desired output and the real output value that represent respectively p sample;

System to the total error criteria function of P training sample is:

$E=\frac{1}{2}\underset{p=1}{\overset{P}{\mathrm{\Σ}}}\underset{k=1}{\overset{L}{\mathrm{\Σ}}}{({T_k}^p-{O_k}^p)}^2$

T wherein _k ^pand O _k ^pthe desired output and the real output value that represent respectively p sample;

According to error gradient descent method, revise successively the correction amount w of output layer weights _ki, the correction amount a of output layer threshold value _k, the correction amount w of hidden layer weights _ij, the correction amount θ of hidden layer threshold value _i;

${\mathrm{\Δw}}_{\mathrm{ki}}=\mathrm{\η}\underset{p=1}{\overset{P}{\mathrm{\Σ}}}\underset{k=1}{\overset{L}{\mathrm{\Σ}}}({T_k}^p-{o_k}^p)\·{\mathrm{\ψ}}^{\′}\left({\mathrm{net}}_k\right)\·y_i$

${\mathrm{\Δa}}_k=\mathrm{\η}\underset{p=1}{\overset{P}{\mathrm{\Σ}}}\underset{k=1}{\overset{L}{\mathrm{\Σ}}}({T_k}^p-{o_k}^p)\·{\mathrm{\ψ}}^{\′}\left({\mathrm{net}}_k\right)$

${\mathrm{\Δw}}_{\mathrm{ij}}=\mathrm{\η}\underset{p=1}{\overset{P}{\mathrm{\Σ}}}\underset{k=1}{\overset{L}{\mathrm{\Σ}}}(T_k^p-o_k^p)\·{\mathrm{\ψ}}^{\′}\left({\mathrm{net}}_k\right)\·w_{\mathrm{ki}}\·{\mathrm{\φ}}^{\′}\left({\mathrm{net}}_i\right)\·x_j$

${\mathrm{\Δ\θ}}_i=\mathrm{\η}\underset{p=1}{\overset{P}{\mathrm{\Σ}}}\underset{k=1}{\overset{L}{\mathrm{\Σ}}}(T_k^p-o_k^p)\·{\mathrm{\ψ}}^{\′}\left({\mathrm{net}}_k\right)\·w_{\mathrm{ki}}\·{\mathrm{\φ}}^{\′}\left({\mathrm{net}}_i\right)$

Wherein η is learning rate, and ψ ' (net _k) represent the differentiate function of hidden layer and output layer excitation function;

After error correction completes, choose at random next sample and offer network, double counting process, until network global error function is less than predefined minimal value, i.e. a network convergence; Or study number of times be less than predefined value, network cannot be restrained.

Vehicle headway computing method in described step 4: when there being vehicle to enter successively after zig zag, road traffic detection device records vehicle in zig zag is by time, the speed of a motor vehicle and this link traffic flow, when a upper road traffic pick-up unit detects the correlation parameter of rear car, by the road traffic pick-up unit image data that transfers to center-controlling computer, carry out front and back following distance and calculate and judge front truck motion state, computing method are:

(1) according to direction of traffic, the road traffic pick-up unit of laying is numbered according to natural order, from starting point road traffic pick-up unit 1, when having vehicle passing detection section, by the time of passing through, the speed of a motor vehicle and the volume of traffic in this this moment of section of this car of road traffic detection device records, and be transferred in time center-controlling computer;

(2) if road traffic pick-up unit h (h is the self-heating number being more than or equal to) at t ₁time detecting is passed through to n (n is more than or equal to 1 a natural number) car, and the speed of a motor vehicle is V ₁, now this section part volume of traffic is n, n+1 car detected and pass through after elapsed time Δ t, the speed of a motor vehicle is V ₂, now this section part volume of traffic is n+1, utilizes follow-up road traffic pick-up unit data search t+ Δ t that computer acquisition the arrives position of n car constantly, the result that may occur is as follows:

1. to t+ Δ t moment h+1 road traffic pick-up unit, n car not detected passes through, now the volume of traffic of h+1 road traffic pick-up unit statistics is n-1, illustrate that now n car and n+1 workshop do not meet safe distance, it is carried out to speed warning during the warning device after n+1 car is by road traffic pick-up unit h;

2. t+ Δ t constantly n car by the individual road traffic pick-up unit of follow-up m (m is more than or equal to 1 natural number), the time t when searching for it and passing through last road traffic pick-up unit ₂and the speed V during by latter two road traffic pick-up unit _h+m-1and V _h+m, calculate respectively required safe distance between the actual range of now n car and n+1 car and two cars, in the time that actual range is less than required safe distance, rear car is carried out to speed early warning.

The workshop actual range of n car and n+1 car is calculated as follows:

$S=\mathrm{ms}+V_{h+m}[\mathrm{\Δt}-(t_2-t_1)]+\frac{({V_{h+m}}^2-{V_{h+m-1}}^2)}{4s}{[\mathrm{\Δt}-(t_2-t_1)]}^2$

In formula

S is road traffic pick-up unit arrangement pitch;

Δ t is the mistiming that road traffic pick-up unit h detects n car and n+1 car;

V _h+mthe speed of a motor vehicle while being n car by h+m road traffic pick-up unit;

V _h+m-1the speed of a motor vehicle while being n car by h+m-1 road traffic pick-up unit;

M is the road traffic pick-up unit number after the road traffic pick-up unit h that passes through within the Δ t time of n car;

T ₁be that n car is by the moment of road traffic pick-up unit h;

T ₂be that n car passes through the moment of last road traffic pick-up unit within the Δ t time;

Now the speed of a motor vehicle of n car is:

${V_2}^{\′}=V_{h+m}+\frac{({V_{h+m}}^2-{V_{h+m-1}}^2)}{2s}[\mathrm{\Δt}-(t_2-t_1)]$

In the time of calculating safe distance required between n+1 car and n car, need first by the BP nerve network controller of structure in the correlation parameter input step 2 of n+1 the car gathering in step 1, road-adhesion coefficient and coefficient of rolling resistance in the hope of n+1 car at this sharp turn.

In described step 5 safe distance calculating: try to achieve n+1 car after the road-adhesion coefficient and coefficient of rolling resistance of sharp turn, safe distance determines that method is:

I) work as V ₂' > V ₂time, vehicle headway meets safe spacing needs, without carrying out vehicle speed prewarning;

II) work as V _h+m< V ₂'≤V ₂time, the safe distance in two workshops is:

$L=\frac{{V_2}^2-V_2^{\&prime;2}}{254(\mathrm{\&phi;}+{\mathrm{\&psi;}}_1)}$

III) work as V ₂'≤V ₂and V ₂'≤V _h+mtime, two shop safety distances are:

$L=\frac{{V_2}^2}{254(\mathrm{\&phi;}+{\mathrm{\&psi;}}_1)}$

In formula and ψ ₁be respectively the n+1 car that calculates according to aforementioned condition at road-adhesion coefficient and the coefficient of rolling resistance of sharp turn.

Described step 6 speed of a motor vehicle is reported to the police and is realized: before each the road traffic pick-up unit arranging on zig zag highway, buzzing warning device is set, when not meeting safe distance requirement by above-mentioned calculative determination rear car and leading vehicle distance, when travelling to warning device place, rear car command it to send speed of a motor vehicle alarm by center-controlling computer.

The invention has the beneficial effects as follows:

The present invention propose a kind of can be according to real-time meteorology, transportation condition, in conjunction with highway layout parameter, the early warning system of reminding while automobile being travelled to hypervelocity in section, sharp turn, so that vehicle is safer when by section, sharp turn, not only more accurate than traditional bend vehicle speed prewarning method, and while being more applicable for inclement weathers such as occurring dense fog, dust storm, sleet, contribute to prevent the generation of traffic hazard.

Accompanying drawing explanation

Figure 1B P neural network structure;

Fig. 2 is at the laying schematic diagram of highway division road traffic pick-up unit;

Fig. 3 workflow diagram of the present invention;

Embodiment

Below in conjunction with accompanying drawing, the present invention is described in detail:

The present invention includes following steps:

1. lay basic information collection device: comprise two kinds of on-vehicle information harvester and road traffic pick-up units.On-vehicle information harvester is by infrared temperature, humidity sensor being set at automobile chassis and saw sensor being set on tire, and collection vehicle enters before highway zig zag within the scope of 20-50m the tire pressure of temperature, humidity and the tire on road surface, deflection, contact area, slippage situation over time; Road traffic pick-up unit mainly gathers the speed of a motor vehicle, the volume of traffic and vehicle and passes through temporal information;

2. based on BP neural network, build the computing controller of road-adhesion coefficient and coefficient of rolling resistance: in truck-mounted computer, using the data item of on-vehicle information harvester Real-time Collection as input, using road-adhesion coefficient and coefficient of rolling resistance as output, chamber test data and existing correlation parameter historical data are carried out sample training by experiment, training error is reduced to predetermined threshold or reaches frequency of training, to obtain optimal neural network controller;

3. the data transmission between truck-mounted computer and center-controlling computer: vehicle is entered to the data input BP nerve network controller that on-vehicle information harvester gathers before zig zag, attachment coefficient and the rolling resistance coefficient of output vehicle on this zig zag, and transfer to trackside RSU equipment and then transfer to center-controlling computer by vehicle-mounted OBU;

4. vehicle headway calculates: when vehicle enters after zig zag successively, road traffic detection device records vehicle in zig zag is by time, the speed of a motor vehicle and this link traffic flow, and by real-time data transmission to center-controlling computer, the front truck speed of a motor vehicle that center-controlling computer detects according to road traffic pick-up unit, front truck by time, the road traffic pick-up unit of each road traffic pick-up unit lay spacing and the rear car speed of a motor vehicle, rear car current location is calculated vehicle actual pitch and front vehicle speed and differentiates the current motion state of front truck;

5. Calculation of Safety Distance: center-controlling computer is according to the safe distance in two workshops before and after the calculation of design parameters of the highway zig zag of the information of vehicles of road traffic pick-up unit collection, the data of rear car truck-mounted computer transmission, the front truck speed of a motor vehicle of having tried to achieve and storage in advance;

6. the speed of a motor vehicle is reported to the police and is realized: before each the road traffic pick-up unit arranging on zig zag highway, buzzing warning device is set, when not meeting safe distance requirement by above-mentioned calculative determination rear car and leading vehicle distance, when rear car is travelled to warning device place, it is sent to speed of a motor vehicle alarm.

Take a sudden turn in described step 1 setting of road traffic pick-up unit on highway: from the starting point of zig zag, consider highway layout vehicle speed value, weather and the pavement conditions average safe stopping distance value of vehicle when better of take is that spacing normative reference is arranged road traffic pick-up unit, in order to make road traffic pick-up unit when work as far as possible not affected by environment, select coil road traffic pick-up unit or microwave road traffic pick-up unit.

The computing formula of vehicle safe braking distance is:

$D=\frac{V^2}{254(\mathrm{\&phi;}+{\mathrm{\&psi;}}_1)}$

In formula, D is vehicle safe braking distance, and V is the average speed of typical vehicle when entering this zig zag on road, and ψ ₁wei weather and pavement conditions attachment coefficient and the road resistance coefficient between road surface and tire, wherein ψ when better ₁=f+i _z, f is coefficient of rolling resistance, i _zfor road longitudinal grade.

In described step 2, BP neural network builds coefficient of road adhesion and coefficient of rolling resistance computing controller is divided into three-decker, i.e. input layer, and hidden layer and output layer, concrete structure is shown in Fig. 1.

Ground floor is input layer, and input value is expressed as x={x ₁, x ₂... x _m, wherein, M represents the number of input parameter, M is more than or equal to 1 integer, x ₁, x ₂... x _mrepresent respectively vehicle, the speed of a motor vehicle, tire pressure, deflection, contact area, slippage, pavement temperature and humidity; x _jrepresent x ₁, x ₂... x _many one parameter, i.e. the arbitrary parameter of input layer;

The second layer is hidden layer, the arbitrary node i of hidden layer be input as net _i

${\mathrm{net}}_i=\underset{j=1}{\overset{M}{\mathrm{\&Sigma;}}}{w}_{\mathrm{ij}}{x}_j+{\mathrm{\&theta;}}_i,$

I node is output as y _i

$y_i=\mathrm{\&phi;}\left({\mathrm{net}}_i\right)=\mathrm{\&phi;}(\underset{j=1}{\overset{M}{\mathrm{\&Sigma;}}}{w}_{\mathrm{ij}}{x}_j+{\mathrm{\&theta;}}_i)$

Wherein

W _ijrepresent that j node of input layer is to the weights of i node of hidden layer, wherein j and i represent respectively the numbering of input layer and the arbitrary node of hidden layer, j=1, and 2,3...M, i=1,2,3...q, M and q are respectively the node number of input layer and hidden layer;

θ _ithe threshold value that represents i node of hidden layer;

φ (x) represents the excitation function of hidden layer;

The 3rd layer is output layer, finally obtains coefficient of road adhesion and coefficient of rolling resistance value, k node of output layer be input as net _k

${\mathrm{net}}_k=\underset{i=1}{\overset{q}{\mathrm{\&Sigma;}}}{w}_{\mathrm{ki}}{y}_i+a_k=\underset{i=1}{\overset{q}{\mathrm{\&Sigma;}}}{w}_{\mathrm{ki}}\mathrm{\&phi;}(\underset{j=1}{\overset{M}{\mathrm{\&Sigma;}}}{w}_{\mathrm{ij}}{x}_j+{\mathrm{\&theta;}}_i)+a_k$

K node is output as o _k

$o_k=\mathrm{\&psi;}\left({\mathrm{net}}_k\right)=\mathrm{\&psi;}(\underset{i=1}{\overset{q}{\mathrm{\&Sigma;}}}{w}_{\mathrm{ki}}{y}_i+a_k)=\mathrm{\&psi;}(\underset{i=1}{\overset{q}{\mathrm{\&Sigma;}}}{w}_{\mathrm{ki}}\mathrm{\&phi;}(\underset{j=1}{\overset{M}{\mathrm{\&Sigma;}}}{w}_{\mathrm{ij}}{x}_j+{\mathrm{\&theta;}}_i)+a_k)$

Wherein

W _kirepresent that the arbitrary node i of hidden layer is to the weights of the arbitrary node k of output layer, k is the arbitrary node serial number of output layer, k=1,2,3...L;

A _kthe threshold value that represents k node of output layer;

ψ (x) represents the excitation function of output layer;

O _krepresent the output of k node of output layer;

During sample training, according to the backpropagation of error, first by output layer, start successively to calculate each layer of neuronic output error, then according to error gradient descent method, regulate weights and the threshold value of each layer, make the final output of amended network can approach expectation value;

Input P learning sample, use x ¹, x ²... x ^p... x ^prepresent, for p sample x ^pquadratic form error rule function be E _p:

$E_p=\frac{1}{2}\underset{k=1}{\overset{L}{\mathrm{\&Sigma;}}}{({T_k}^p-{O_k}^p)}^2$

T wherein _kit is the desired output of k node;

System to the total error criteria function of P training sample is:

$E=\frac{1}{2}\underset{p=1}{\overset{P}{\mathrm{\&Sigma;}}}\underset{k=1}{\overset{L}{\mathrm{\&Sigma;}}}{({T_k}^p-{O_k}^p)}^2$

T wherein _k ^pand O _k ^pthe desired output and the real output value that represent respectively p sample;

According to error gradient descent method, revise successively the correction amount w of output layer weights _ki, the correction amount a of output layer threshold value _k, the correction amount w of hidden layer weights _ij, the correction amount θ of hidden layer threshold value _i;

${\mathrm{\&Delta;w}}_{\mathrm{ki}}=\mathrm{\&eta;}\underset{p=1}{\overset{P}{\mathrm{\&Sigma;}}}\underset{k=1}{\overset{L}{\mathrm{\&Sigma;}}}({T_k}^p-{o_k}^p)\&CenterDot;{\mathrm{\&psi;}}^{\&prime;}\left({\mathrm{net}}_k\right)\&CenterDot;y_i$

${\mathrm{\&Delta;a}}_k=\mathrm{\&eta;}\underset{p=1}{\overset{P}{\mathrm{\&Sigma;}}}\underset{k=1}{\overset{L}{\mathrm{\&Sigma;}}}({T_k}^p-{o_k}^p)\&CenterDot;{\mathrm{\&psi;}}^{\&prime;}\left({\mathrm{net}}_k\right)$

${\mathrm{\&Delta;w}}_{\mathrm{ij}}=\mathrm{\&eta;}\underset{p=1}{\overset{P}{\mathrm{\&Sigma;}}}\underset{k=1}{\overset{L}{\mathrm{\&Sigma;}}}(T_k^p-o_k^p)\&CenterDot;{\mathrm{\&psi;}}^{\&prime;}\left({\mathrm{net}}_k\right)\&CenterDot;w_{\mathrm{ki}}\&CenterDot;{\mathrm{\&phi;}}^{\&prime;}\left({\mathrm{net}}_i\right)\&CenterDot;x_j$

${\mathrm{\&Delta;\&theta;}}_i=\mathrm{\&eta;}\underset{p=1}{\overset{P}{\mathrm{\&Sigma;}}}\underset{k=1}{\overset{L}{\mathrm{\&Sigma;}}}(T_k^p-o_k^p)\&CenterDot;{\mathrm{\&psi;}}^{\&prime;}\left({\mathrm{net}}_k\right)\&CenterDot;w_{\mathrm{ki}}\&CenterDot;{\mathrm{\&phi;}}^{\&prime;}\left({\mathrm{net}}_i\right)$

Wherein η is learning rate, φ ' (x) and ψ ' (x) represent the differentiate function of hidden layer and output layer excitation function;

After error correction completes, choose at random next sample and offer network, double counting process, until network global error function is less than predefined minimal value, i.e. a network convergence; Or study number of times be less than predefined value, network cannot be restrained.

Vehicle headway computing method in described step 4: when there being vehicle to enter successively after zig zag, road traffic detection device records vehicle in zig zag is by time, the speed of a motor vehicle and this link traffic flow, when a upper road traffic pick-up unit detects the correlation parameter of rear car, by the road traffic pick-up unit image data that transfers to center-controlling computer, carry out front and back following distance and calculate and judge front truck motion state, computing method are:

(1) according to direction of traffic, the road traffic pick-up unit of laying is numbered to 1,2,3......, from starting point road traffic pick-up unit 1, when having vehicle passing detection section, by the time of passing through, the speed of a motor vehicle and the volume of traffic in this this moment of section of this car of road traffic detection device records, and be transferred in time center-controlling computer;

(2) if road traffic pick-up unit h at t ₁time detecting is passed through to n car, and the speed of a motor vehicle is V ₁, now this section part volume of traffic is n, n+1 car detected and pass through after elapsed time Δ t, the speed of a motor vehicle is V ₂, now this section part volume of traffic is n+1, utilizes follow-up road traffic pick-up unit data search t+ Δ t that computer acquisition the arrives position of n car constantly, the result that may occur is as follows:

1. to t+ Δ t moment h+1 road traffic pick-up unit, n car not detected passes through, now the volume of traffic of h+1 road traffic pick-up unit statistics is n-1, illustrate that now n car and n+1 workshop do not meet safe stopping distance, it is carried out to speed warning during the warning device after n+1 car is by road traffic pick-up unit h;

2. at t+ Δ t moment n car, follow-up m road traffic pick-up unit, the time t while searching for it by last road traffic pick-up unit have been passed through ₂and the speed V during by latter two road traffic pick-up unit _h+m-1and V _h+m, calculate respectively required safe distance between the actual range of now n car and n+1 car and two cars, in the time that actual range is less than required safe distance, rear car is carried out to speed early warning.

The workshop actual range of n car and n+1 car is calculated as follows:

$S=\mathrm{ms}+V_{h+m}[\mathrm{\&Delta;t}-(t_2-t_1)]+\frac{({V_{h+m}}^2-{V_{h+m-1}}^2)}{4s}{[\mathrm{\&Delta;t}-(t_2-t_1)]}^2$

In formula

S is road traffic pick-up unit arrangement pitch;

Δ t is the mistiming that road traffic pick-up unit i detects n car and n+1 car;

V _h+mthe speed of a motor vehicle while being n car by h+m road traffic pick-up unit;

V _h+m-1the speed of a motor vehicle while being n car by h+m-1 road traffic pick-up unit;

M is the road traffic pick-up unit number after the road traffic pick-up unit h that passes through within the Δ t time of n car;

T ₁be that n car is by the moment of road traffic pick-up unit h;

T ₂be that n car passes through the moment of last road traffic pick-up unit within the Δ t time;

Now the speed of a motor vehicle of n car is:

${V_2}^{\&prime;}=V_{h+m}+\frac{({V_{h+m}}^2-{V_{h+m-1}}^2)}{2s}[\mathrm{\&Delta;t}-(t_2-t_1)]$

In the time of calculating safe distance required between n+1 car and n car, need first by the BP nerve network controller of structure in the correlation parameter input step 2 of n+1 the car gathering in step 1, road-adhesion coefficient and coefficient of rolling resistance in the hope of n+1 car at this sharp turn.

In described step 5 safe distance calculating: try to achieve n+1 car after the road-adhesion coefficient and coefficient of rolling resistance of sharp turn, safe distance determines that method is:

I) work as V ₂' > V ₂time, vehicle headway meets safe spacing needs, without carrying out vehicle speed prewarning;

II) work as V _h+m< V ₂'≤V ₂time, the safe distance in two workshops is:

$L=\frac{{V_2}^2-V_2^{\&prime;2}}{254(\mathrm{\&phi;}+{\mathrm{\&psi;}}_1)}$

III) work as V ₂'≤V ₂and V ₂'≤V _h+mtime, two shop safety distances are:

$L=\frac{{V_2}^2}{254(\mathrm{\&phi;}+{\mathrm{\&psi;}}_1)}$

In formula and ψ ₁be respectively the n+1 car that calculates according to aforementioned condition at road-adhesion coefficient and the coefficient of rolling resistance of sharp turn.

Described step 6 speed of a motor vehicle is reported to the police and is realized: before each the road traffic pick-up unit arranging in zig zag, buzzing warning device is set, when not meeting safe distance requirement by above-mentioned calculative determination rear car and leading vehicle distance, when travelling to warning device place, rear car command it to send speed of a motor vehicle alarm by center-controlling computer.

Claims (7)

1. the real-time vehicle speed prewarning method in section, mountain highway sharp turn, is characterized in that, comprises the steps:

Step 1. is laid basic information collection device on vehicle and zig zag highway, gathers road and information of vehicles on the spot, and described basic information collection device comprises on-vehicle information harvester and road road traffic pick-up unit;

Step 2 builds the computing controller of road-adhesion coefficient and coefficient of rolling resistance in truck-mounted computer based on BP neural network;

The input information BP nerve network controller that step 3 gathers on-vehicle information harvester in step 1 is processed, export vehicle at road-adhesion coefficient and the coefficient of rolling resistance of sharp turn, and transfer to trackside microwave read-write antenna equipment and then transfer to center-controlling computer by vehicle-carried microwave communication apparatus;

The front truck speed of a motor vehicle that step 4 center-controlling computer detects according to road traffic pick-up unit, front truck are laid spacing and the rear car speed of a motor vehicle, rear car current location calculating vehicle actual pitch and are differentiated the current motion state of front truck by time, the road traffic pick-up unit of each road traffic pick-up unit;

The signal of the real-time speed of a motor vehicle of rear car that step 5 center-controlling computer detects according to road traffic pick-up unit, step (3) output and the in advance current section highway layout parameter of storage, obtain the safe distance between vehicle;

Safe distance and vehicle actual pitch between step 6 pair vehicle compare, if vehicle actual pitch is less than safe distance, the rear car speed of a motor vehicle are reported to the police, if vehicle actual pitch is greater than safe distance, get back to the signal that step 1 gathers next car.

2. the real-time vehicle speed prewarning method in section, a kind of mountain highway sharp turn as claimed in claim 1, it is characterized in that, in described step 1, on-vehicle information harvester is saw sensor and infrared temperature, humidity sensor, described sensor is located on vehicle chassis and tire, and collection vehicle enters take a sudden turn tire pressure, deflection, contact area, slippage and pavement temperature, the humidity parameter situation over time of tire within the scope of about 20-50m of highway; Described road is layed on road with road traffic pick-up unit, gathers the speed of a motor vehicle, the volume of traffic and vehicle and passes through temporal information; The data that on-vehicle information harvester gathers are passed through trackside wireless telecommunications system real-time Transmission to center-controlling computer after truck-mounted computer is processed, and the information of road traffic pick-up unit collection is transmitted through the fiber to center-controlling computer.

3. the real-time vehicle speed prewarning method in section, a kind of mountain highway sharp turn as claimed in claim 1, it is characterized in that, in described step 1, road traffic pick-up unit method to set up on zig zag highway: from zig zag starting point, the average safe stopping distance value of vehicle of take while considering that under highway layout speed of a motor vehicle condition, weather and pavement conditions are better is standard, every the average safe stopping distance of vehicle, establish a road traffic pick-up unit, to realize real-time detection and the transmission of the volume of traffic and vehicle speed data on road; In order to make road traffic pick-up unit when work as far as possible not affected by environment, select coil road traffic pick-up unit or microwave road traffic pick-up unit.

The computing formula of vehicle safe braking distance is:

$D=\frac{V^2}{254(\mathrm{\&phi;}+{\mathrm{\&psi;}}_1)}$

In formula, D is vehicle safe braking distance, and V is the average speed of typical vehicle when entering this zig zag on road, and ψ ₁wei weather and pavement conditions attachment coefficient and the road resistance coefficient between road surface and tire, wherein ψ when better ₁=f+i _z, f is coefficient of rolling resistance, i _zfor road longitudinal grade.

4. the real-time vehicle speed prewarning method in section, a kind of mountain highway sharp turn as claimed in claim 1, it is characterized in that, in described step 2, the process of establishing of described computing controller is: using the data item of on-vehicle information harvester Real-time Collection in step 1 as input, using road-adhesion coefficient and coefficient of rolling resistance as output, chamber test data and existing correlation parameter historical data are carried out sample training by experiment, training error is reduced to predetermined threshold or reaches frequency of training, to obtain optimal neural network controller.

5. the real-time vehicle speed prewarning method in section, a kind of mountain highway sharp turn as claimed in claim 1, is characterized in that, in described step 2, computing controller is divided into three-decker, i.e. input layer, hidden layer and output layer;

Ground floor is input layer, and input value is expressed as x={x ₁, x ₂... x _m, wherein, M represents the number of input parameter, M is more than or equal to 1 natural number, x ₁, x ₂... x _mrepresent respectively vehicle, the speed of a motor vehicle, tire pressure, deflection, contact area, slippage, pavement temperature and humidity; x _jrepresent x ₁, x ₂... x _min any one parameter, i.e. the arbitrary parameter of input layer;

The second layer is hidden layer, the arbitrary node i of hidden layer be input as net _i

${\mathrm{net}}_i=\underset{j=1}{\overset{M}{\mathrm{\&Sigma;}}}{w}_{\mathrm{ij}}{x}_j+{\mathrm{\&theta;}}_i,$

I node is output as y _i

$y_i=\mathrm{\&phi;}\left({\mathrm{net}}_i\right)=\mathrm{\&phi;}(\underset{j=1}{\overset{M}{\mathrm{\&Sigma;}}}{w}_{\mathrm{ij}}{x}_j+{\mathrm{\&theta;}}_i)$

Wherein

W _ijrepresent that j node of input layer is to the weights of i node of hidden layer, wherein j and i represent respectively the numbering of input layer and the arbitrary node of hidden layer, j=1,2,3...M, i=1,2,3...q, M and q are respectively the node number of input layer and hidden layer, and M, and q is more than or equal to 1 natural number;

θ _ithe threshold value that represents i node of hidden layer;

φ (x) represents the excitation function of hidden layer;

The 3rd layer is output layer, finally obtains coefficient of road adhesion and coefficient of rolling resistance value, k node of output layer be input as net _k

${\mathrm{net}}_k=\underset{i=1}{\overset{q}{\mathrm{\&Sigma;}}}{w}_{\mathrm{ki}}{y}_i+a_k=\underset{i=1}{\overset{q}{\mathrm{\&Sigma;}}}{w}_{\mathrm{ki}}\mathrm{\&phi;}(\underset{j=1}{\overset{M}{\mathrm{\&Sigma;}}}{w}_{\mathrm{ij}}{x}_j+{\mathrm{\&theta;}}_i)+a_k$

K node is output as o _k

$o_k=\mathrm{\&psi;}\left({\mathrm{net}}_k\right)=\mathrm{\&psi;}(\underset{i=1}{\overset{q}{\mathrm{\&Sigma;}}}{w}_{\mathrm{ki}}{y}_i+a_k)=\mathrm{\&psi;}(\underset{i=1}{\overset{q}{\mathrm{\&Sigma;}}}{w}_{\mathrm{ki}}\mathrm{\&phi;}(\underset{j=1}{\overset{M}{\mathrm{\&Sigma;}}}{w}_{\mathrm{ij}}{x}_j+{\mathrm{\&theta;}}_i)+a_k)$

Wherein

W _kirepresent that the arbitrary node i of hidden layer is to the weights of the arbitrary node k of output layer, k is the arbitrary node serial number of output layer, k=1, and 2,3...L, L is more than or equal to 1 natural number;

A _kthe threshold value that represents k node of output layer;

Ψ (net _k) represent the excitation function of output layer;

O _krepresent the output of k node of output layer;

During sample training, according to the backpropagation of error, first by output layer, start successively to calculate each layer of neuronic output error, then according to error gradient descent method, regulate weights and the threshold value of each layer, make the final output of amended network can approach expectation value;

Input P learning sample, use x ¹, x ²... x ^p, representing, described p is more than or equal to 1 natural number;

For p sample x ^pquadratic form error rule function be E _p:

$E_p=\frac{1}{2}\underset{k=1}{\overset{L}{\mathrm{\&Sigma;}}}{({T_k}^p-{O_k}^p)}^2$

System to the total error criteria function of P training sample is:

$E=\frac{1}{2}\underset{p=1}{\overset{P}{\mathrm{\&Sigma;}}}\underset{k=1}{\overset{L}{\mathrm{\&Sigma;}}}{({T_k}^p-{O_k}^p)}^2$

T wherein _k ^pand O _k ^pthe desired output and the real output value that represent respectively p sample;

According to error gradient descent method, revise successively the correction amount w of output layer weights _ki, the correction amount a of output layer threshold value _k, the correction amount w of hidden layer weights _ij, the correction amount θ of hidden layer threshold value _i;

${\mathrm{\&Delta;w}}_{\mathrm{ki}}=\mathrm{\&eta;}\underset{p=1}{\overset{P}{\mathrm{\&Sigma;}}}\underset{k=1}{\overset{L}{\mathrm{\&Sigma;}}}({T_k}^p-{o_k}^p)\&CenterDot;{\mathrm{\&psi;}}^{\&prime;}\left({\mathrm{net}}_k\right)\&CenterDot;y_i$

${\mathrm{\&Delta;a}}_k=\mathrm{\&eta;}\underset{p=1}{\overset{P}{\mathrm{\&Sigma;}}}\underset{k=1}{\overset{L}{\mathrm{\&Sigma;}}}({T_k}^p-{o_k}^p)\&CenterDot;{\mathrm{\&psi;}}^{\&prime;}\left({\mathrm{net}}_k\right)$

${\mathrm{\&Delta;w}}_{\mathrm{ij}}=\mathrm{\&eta;}\underset{p=1}{\overset{P}{\mathrm{\&Sigma;}}}\underset{k=1}{\overset{L}{\mathrm{\&Sigma;}}}(T_k^p-o_k^p)\&CenterDot;{\mathrm{\&psi;}}^{\&prime;}\left({\mathrm{net}}_k\right)\&CenterDot;w_{\mathrm{ki}}\&CenterDot;{\mathrm{\&phi;}}^{\&prime;}\left({\mathrm{net}}_i\right)\&CenterDot;x_j$

${\mathrm{\&Delta;\&theta;}}_i=\mathrm{\&eta;}\underset{p=1}{\overset{P}{\mathrm{\&Sigma;}}}\underset{k=1}{\overset{L}{\mathrm{\&Sigma;}}}(T_k^p-o_k^p)\&CenterDot;{\mathrm{\&psi;}}^{\&prime;}\left({\mathrm{net}}_k\right)\&CenterDot;w_{\mathrm{ki}}\&CenterDot;{\mathrm{\&phi;}}^{\&prime;}\left({\mathrm{net}}_i\right)$

Wherein η is learning rate, and ψ ' (net _k) represent the differentiate function of hidden layer and output layer excitation function;

After error correction completes, choose at random next sample and offer network, double counting process, until network global error function is less than predefined minimal value, i.e. a network convergence; Or study number of times be less than predefined value, network cannot be restrained.

6. the real-time vehicle speed prewarning method in section, a kind of mountain highway sharp turn as claimed in claim 1, it is characterized in that, described step 4 vehicle headway computing method: when there being vehicle to enter successively after zig zag, road traffic detection device records vehicle in zig zag is by time, the speed of a motor vehicle and this link traffic flow, when a upper road traffic pick-up unit detects the correlation parameter of rear car, by the road traffic pick-up unit image data that transfers to center-controlling computer, carry out front and back following distance and calculate and judge front truck motion state, computing method are:

(1) according to direction of traffic, the road traffic pick-up unit of laying is numbered according to natural order, from starting point road traffic pick-up unit 1, when having vehicle passing detection section, by the time of passing through, the speed of a motor vehicle and the volume of traffic in this this moment of section of this car of road traffic detection device records, and be transferred in time center-controlling computer;

(2) if road traffic pick-up unit h at t ₁time detecting is passed through to n car, and the speed of a motor vehicle is V ₁, now this section part volume of traffic is n, n+1 car detected and pass through after elapsed time Δ t, the speed of a motor vehicle is V ₂, now this section part volume of traffic is n+1, utilizes follow-up road traffic pick-up unit data search t+ Δ t that computer acquisition the arrives position of n car constantly, the result that may occur is as follows:

1. to t+ Δ t moment h+1 road traffic pick-up unit, n car not detected passes through, now the volume of traffic of h+1 road traffic pick-up unit statistics is n-1, illustrate that now n car and n+1 workshop do not meet safe stopping distance, it is carried out to speed warning during the warning device after n+1 car is by road traffic pick-up unit h;

2. at t+ Δ t moment n car, follow-up m road traffic pick-up unit, the time t while searching for it by last road traffic pick-up unit have been passed through ₂and the speed V during by latter two road traffic pick-up unit _h+m-1and V _h+m, calculate respectively required safe distance between the actual range of now n car and n+1 car and two cars, in the time that actual range is less than required safe distance, rear car is carried out to speed early warning.

The workshop actual range of n car and n+1 car is calculated as follows:

$S=\mathrm{ms}+V_{h+m}[\mathrm{\&Delta;t}-(t_2-t_1)]+\frac{({V_{h+m}}^2-{V_{h+m-1}}^2)}{4s}{[\mathrm{\&Delta;t}-(t_2-t_1)]}^2$

In formula

S is road traffic pick-up unit arrangement pitch;

Δ t is the mistiming that road traffic pick-up unit h detects n car and n+1 car;

V _h+mthe speed of a motor vehicle while being n car by h+m road traffic pick-up unit;

V _h+m-1the speed of a motor vehicle while being n car by h+m-1 road traffic pick-up unit;

M is the road traffic pick-up unit number after the road traffic pick-up unit h that passes through within the Δ t time of n car;

T ₁be that n car is by the moment of road traffic pick-up unit h;

T ₂be that n car passes through the moment of last road traffic pick-up unit within the Δ t time;

Now the speed of a motor vehicle of n car is:

${V_2}^{\&prime;}=V_{h+m}+\frac{({V_{h+m}}^2-{V_{h+m-1}}^2)}{2s}[\mathrm{\&Delta;t}-(t_2-t_1)]$

In the time of calculating safe distance required between n+1 car and n car, need first by the BP nerve network controller of structure in the correlation parameter input step 2 of n+1 the car gathering in step 1, road-adhesion coefficient and coefficient of rolling resistance in the hope of n+1 car at this sharp turn.

In described step 5 safe distance calculating: try to achieve n+1 car after the road-adhesion coefficient and coefficient of rolling resistance of sharp turn, safe distance determines that method is:

I) work as V ₂' > V ₂time, vehicle headway meets safe spacing needs, without carrying out vehicle speed prewarning;

II) work as V _h+m< V ₂'≤V ₂time, the safe distance in two workshops is:

$L=\frac{{V_2}^2-V_2^{\&prime;2}}{254(\mathrm{\&phi;}+{\mathrm{\&psi;}}_1)}$

III) work as V ₂'≤V ₂and V ₂'≤V _h+mtime, two shop safety distances are:

$L=\frac{{V_2}^2}{254(\mathrm{\&phi;}+{\mathrm{\&psi;}}_1)}$

In formula and ψ ₁be respectively the n+1 car that calculates according to aforementioned condition at road-adhesion coefficient and the coefficient of rolling resistance of sharp turn.

7. the real-time vehicle speed prewarning method in section, a kind of mountain highway sharp turn as claimed in claim 1, it is characterized in that, speed of a motor vehicle warning implementation in described step 6: before each the road traffic pick-up unit arranging, buzzing warning device is set in zig zag, when not meeting safe distance requirement by above-mentioned calculative determination vehicle headway, when rear car is travelled to warning device place, it is sent to speed of a motor vehicle alarm.